agentsclimarketplace

Sre runbook executor

Skill medzin/sre-runbook-agent-skills/skills/sre-runbook-executor

Agent Skills for writing, reviewing, and safely executing production SRE runbooks.

Install
npx -y skills add medzin/sre-runbook-agent-skills --skill sre-runbook-executor

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Execute production SRE runbooks step by step during incidents or operational tasks. Use when the user asks an agent to follow, run, execute, apply, or walk through a runbook safely with read-only diagnostics first, evidence capture, verification, ambiguity stops, escalation, and approval gates for risky actions.

SKILL.md

2.3 KB, as published. Nobody here has run it

SRE Runbook Executor

Use this skill when the user asks to execute, follow, apply, or walk through an existing production SRE runbook.

Do not use this skill to create a new runbook or to improvise missing operational procedures. If the runbook is incomplete, stop and ask for the missing detail.

Core Behavior

  • Restate the incident or input.
  • Identify the runbook being used and why it applies.
  • Execute only one step at a time.
  • Prefer read-only diagnostics first.
  • Use available MCP servers, apps, or connectors only for read-only evidence collection and context verification for the current step.
  • Record evidence for every step.
  • Compare actual results with expected results.
  • Stop if a step is ambiguous.
  • Stop if required access or tools are missing.
  • Stop if observed state does not match the runbook.
  • Never invent missing commands, thresholds, owners, dashboards, or expected outputs.
  • Ask for approval before any risky action.
  • Never use connector data to silently skip an approval gate, broaden scope, or perform a mutating action not explicitly present in the runbook.

Risky Actions

Risky actions include:

  • Destructive commands
  • Customer-impacting changes
  • Data changes
  • Deploys
  • Rollbacks
  • Restarts
  • Scaling operations
  • Failovers
  • Config changes
  • Permission changes
  • Disabling alerts
  • Deleting or modifying resources

References

Open these only when useful:

  • references/execution-safety-contract.md for the execution rules.
  • references/approval-gates.md for approval requirements.
  • references/execution-log-template.md for evidence logging.
  • references/context-evidence.md when read-only connector lookups can support execution evidence.

Execution Output Format

For every step, use:

  • Current step
  • Action to perform
  • Safety classification: read-only / low-risk / risky
  • Required approval: yes/no
  • Evidence collected
  • Expected result
  • Actual result
  • Decision
  • Next step
  • Stop condition, if any

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.